Comparison of Commensal and Clinical Isolates for Diversity of Plasmids in Escherichia coli and Klebsiella pneumoniae

In this study, the plasmid content of clinical and commensal strains was analyzed and compared. The replicon profile was similar in both populations, except for L, M, A/C, and N (detected only in clinical strains) and HI1 (only in commensal strains). Although I1 and F were the most frequent replicons, only IncI1, sequence type 12 (ST12) was associated with bla_CMY-2 in both populations. In contrast, the widespread resistant IncF plasmids were not linked to a single epidemic plasmid.

ABSTRACT

In this study, the plasmid content of clinical and commensal strains was analyzed and compared. The replicon profile was similar in both populations, except for L, M, A/C, and N (detected only in clinical strains) and HI1 (only in commensal strains). Although I1 and F were the most frequent replicons, only IncI1, sequence type 12 (ST12) was associated with bla_CMY-2 in both populations. In contrast, the widespread resistant IncF plasmids were not linked to a single epidemic plasmid.

TEXT

The most prevalent mechanism in antimicrobial resistance gene (ARG) acquisition by bacterial pathogens is horizontal gene transfer by plasmids (1, 2). PCR-based replicon typing (PBRT) based on plasmid incompatibility (Inc) is currently the standard method for plasmid identification (3, 4). Plasmid multilocus sequence typing (pMLST) schemes allow researchers to differentiate between plasmids within incompatibility groups and to define epidemiological and evolutionary relatedness (5–10) (http://pubmlst.org/plasmid/).

Several plasmids carrying ARGs have been characterized, most of them recovered from clinically relevant bacteria (11–14). In contrast, there is limited information on plasmids in the commensal microbiota of healthy humans without a selection bias for antimicrobial-resistant bacteria. In this scenario, the aim of this study was to provide a better understanding of resistant plasmid diffusion in a clinical context by comparing plasmids within Escherichia coli and Klebsiella pneumoniae strains isolated from healthy human feces and patients with bloodstream infection.

One hundred and fifty fecal samples were collected during 2014 and 2015 from healthy humans who did not consume antibiotics and were not hospitalized. A total of 145 E. coli strains and 12 K. pneumoniae strains were isolated. In addition, 202 strains from blood cultures, 99 E. coli strains and 103 K. pneumoniae strains, from three hospitals in Barcelona were analyzed (one per patient). (The study was approved by the Hospital de la Santa Creu i Sant Pau ethics committee [13/051/1439].) All strains underwent antimicrobial susceptibility testing using disk diffusion (see Table S1 in the supplemental material), and the results were interpreted according to CLSI guidelines (15). The characterization of extended-spectrum β-lactamases (ESBLs), AmpCs, and carbapenemases (16–20) detected in both populations is shown in Table 1. The prevalence of ESBL-producing E. coli in healthy carriers (4.7%) was higher than that in a previous study in Barcelona in 2005 (3.3%) (21) but still within the 3 to 6% average of Europe (22).

TABLE 1.

ESBL, AmpC, and carbapenemase genes detected in E. coli and K. pneumoniae from fecal and blood samples

Gene type (no.)	Gene (no.) detected in:
	E. coli (n = 244)		K. pneumoniae (n = 115)
	Fecal samples (n = 13/145)a	Blood cultures (n = 19/99)b	Fecal samples (n = 0/12)	Blood cultures (n = 17/103)c
ESBLs (43)	n = 8; 5.5%	n = 17; 17.2%		n = 17; 16.5%
	blaCTX-M-15 (4)	blaCTX-M-15 (10)		blaCTX-M-15 (9)
	blaCTX-M-14 (3)	blaCTX-M-14 (2)		blaCTX-M-14 (2)
	blaCTX-M-27 (1)	blaCTX-M-27 (2)		blaSHV-28 (5)
	blaSHV-12 (1)	blaCTX-M-32 (1)		blaSHV-2 (1)
		blaSHV-12 (2)
AmpCs (11)	n = 5; 3.5%	n = 5; 5.0%		n = 1; 1.0%
	blaCMY-2 (5)	blaCMY-2 (4)		blaDHA-1 (1)
		blaDHA-1 (1)
Carbapenemases (2)				n = 2; 1.9%
				blaKPC-3 (2)
Total genes (56)	14	22	0	20

^aOne E. coli strain had bla_CTX-M-15 and bla_CTX-M-14.

^bThree E. coli strains had bla_CTX-M-15 and bla_CTX-M-32, bla_CTX-M-27 and bla_DHA-1, and bla_SHV-12 and bla_CMY-2.

^cOne K. pneumoniae strain had bla_CTX-M-15 and bla_CTX-M-14, and two strains had bla_CTX-M-15 and bla_SHV-28.

Plasmid identification was performed using the PBRT-kit (Diatheva) and simplex PCR for ColE, X3, X4, L, and M replicons (23–25). Twenty-nine replicons were analyzed, and only FIIS, W, T, U, and HI2 were not detected in any strain. A total of 978 replicons were identified in the 359 studied strains; 84.1% (302/359) harbored from one to four replicons, and 10.9% (39/359) harbored from five to seven. In 5% (18/359) of the strains, no replicon was detected. Overall, the results suggest that the replicon content of E. coli strains followed a similar trend in patients and healthy individuals, and the most prevalent in both sample groups were ColE, FII, and FIB (Fig. 1A). Nevertheless, replicons M, A/C, and N were only detected in clinical strains in accordance with the literature (26, 27), while FIIK and HI1 were observed only in fecal strains (Fig. 1A). Hence, it might be hypothesized that the hospital environment, where there is high antimicrobial use and intense interhuman transmission, selects for plasmids more adapted to these settings. The plasmid content in K. pneumoniae isolates seems to follow similar trends (Fig. 1B) to that of E. coli, but this could not be confirmed due to the low number of strains obtained from fecal samples. Notably, both the diversity and frequency of replicons were higher in E. coli than in K. pneumoniae, except for those of R and FIIK (Fig. 1C).

FIG 1.

Replicon prevalence. (a) Comparison between replicons detected in E. coli from clinical and fecal samples. (b) Comparison between replicons detected in K. pneumoniae from clinical and fecal samples. (c) Comparison between replicons detected in E. coli and K. pneumoniae from the total strains. Tables show the number of each replicon detected. ND, not detected. *, Statistical differences (P < 0.05) between each population, fecal and blood samples or E. coli and K. pneumoniae.

IncF and IncI1 plasmids have been reported in Enterobacteriaceae as promoters of beta-lactamase gene dissemination in multiple environments, specifically bla_CTXM-15 and bla_CMY-2 (13, 28–32). In this study, 56 beta-lactamase genes were detected (Table 1). After S1 pulsed-field gel electrophoresis (PFGE) and Southern hybridization (19, 33), 75% of ESBL, AmpC, and carbapenemase genes identified in E. coli (27/36) and K. pneumoniae (15/20) were located on a plasmid, the most prevalent being IncF and IncI1 (37% for both) in E. coli and IncF (47%) and IncR (20%) in K. pneumoniae. The predominant genes were bla_{CTX-M-15/14/27} and bla_CMY-2 in IncF and InI1 plasmids of E. coli (see Fig. S1A in the supplemental material) and bla_CTX-M-15 in IncF plasmids of K. pneumoniae (see Fig. S1B). Figure 2 summarizes the 49 ESBL-, AmpC-, and/or carbapenemase-producing strains detected in the study, the plasmids they harbored, and the location of the beta-lactamase genes.

FIG 2.

Heat-map summary of the sources, phylogenetic groups, β-lactamase resistance genes, and the corresponding Inc plasmid types and their sizes for 49 β-lactam-resistant E. coli (n = 32) and K. pneumoniae (n = 17) strains from fecal and blood samples. Black and white squares denote the presence and absence of a particular feature, respectively. ND, not determined; CP, carbapenemases. ^a, Plasmids where the replicon hybridization occurred in the same plasmid size (hybrid plasmids). ^b, bla Genes detected in the same plasmid.

As IncF and IncI1 were two of the most frequently detected plasmids in both fecal and clinical samples (27, 34, 35), they were further characterized using the pMLST method (5, 10). In E. coli strains, 29 different IncI1 sequence types (STs) were detected, 59% of which were assigned as new STs. This result reflects the great diversity within this plasmid family, with only ST12 and ST36 being present in both clinical and fecal populations (see Fig. S2 in the supplemental material). Moreover, some of the most frequently reported STs worldwide (ST2, ST12, ST26, and ST36) (36) were only found in E. coli from healthy humans. The detection of many newly assigned STs in the clinical isolates and a scarce number of the most reported STs suggest that the latter may have been overreported due to their involvement in ARG dissemination, resulting in an epidemiological bias.

In addition, IncI1 plasmids have been associated with the carriage of bla_CMY-2, particularly IncI1 ST2, ST12, and ST23 (5, 36–38). In the current study, all identified ST12 plasmids harbored bla_CMY-2 and were detected in E. coli from both populations (Fig. 2). These results support the suggestion that some IncI1 plasmids have been able to evolve and persist in clinical settings thanks to particular features that provide resistance, persistence, and adaptive success, which would explain why they are more frequently reported and described as epidemic plasmids (36–38).

After defining the final number of IncF plasmids (n = 279; 211 in E. coli strains and 68 in K. pneumoniae strains) by Southern hybridization of F replicons within each strain, subtypification using replicon sequence typing (RST) was performed. In E. coli strains, 86 different FAB (FII, FIA, FIB) formulas (indicating the allele type and number identified for each replicon according to RST scheme [10]) from 205 typeable plasmids (45 from fecal samples, 24 from blood, and 17 from both) were defined, where F29:A−:B10, F2:A−:B1, F2:A−:B−, and F24:A−:B1 were the most frequent (see Fig. S3 in the supplemental material). Some of these formulas have been previously identified in different environments, such as avian-pathogenic E. coli strains and uropathogenic and extraintestinal pathogenic E. coli strains (27, 39), indicating a broad distribution. In K. pneumoniae, 16 different FAB formulas from 68 typeable plasmids (12 from blood and 4 from both) were detected, with K1:A−:B− being the most frequent (see Fig. S3).

IncF plasmids, including F2:A−:B−, F2:A1:B−, F31:A4:B1, and F1:A2:B20, have been associated with the worldwide emergence of CTX-M-15 (10, 12, 40–42). Although these four plasmids were detected in our study, only F2:A−:B− harbored the bla_CTX-M-15 gene in a clinical strain and bla_CTX-M-27 in a fecal strain. All of the other CTX-M-encoding genes were located in different IncF plasmids (Fig. 2). Thus, according to our results and in agreement with those of previous reports (32, 34), there is no evidence for the persistence of a unique IncF. These highly versatile plasmids are able to adapt to intracellular environments by the rapid evolution of replicon regulatory sequences (10), and they were widely distributed in the Enterobacteriaceae before antimicrobial use, facilitating the persistence and spread of beta-lactamases (10, 32). Their coexistence with other resistance determinants also contributes to the dissemination of IncF-CTX-M plasmids (43).

Additionally, E. coli strains were assigned to phylogenetic groups following the procedure of Clermont et al. (44) (Table 2), and the presence of 15 virulence factors (VFs) was determined (45, 46) (Fig. 3). In commensal E. coli, the prevalence of phylogenetic groups varies among studies (35, 47). It has been reported that the highly diverse hosts and environmental factors, the determinants of virulence, and the antimicrobial pressure can modify prevalence for a better adaptation to commensal habitats (47). In our study, even though commensal E. coli presented a higher diversity of phylogroups compared to that of the clinical samples (Table 2), a predominance of the phylogroup B2 carrying high rates of VFs was found in both populations. Although no evident association has been reported between plasmids and phylogroups (34), our results indicate a possible association of HI1 plasmids with phylogroup A (P ≤ 0.007, Bonferroni’s correction was applied).

TABLE 2.

Phylogenetic groups detected in E. coli strains isolated from fecal and blood samples

E. coli isolate typea	No. (%) detected of each phylogenetic group
	Total no.	A	B1	B2	C	D	E	F	Clade I	Unknown
Total E. coli isolates	244	26 (10.6)	22 (9)	124 (50.8)	3 (1.2)	32 (13.1)	15 (6.1)	17 (6.9)	4 (1.6)	1 (0.41)
Fecal samples	145	23 (15.9)	11 (7.6)	59 (40.7)	3 (2.1)	23 (15.9)	8 (5.5)	13 (8.9)	4 (2.7)	1 (0.7)
Susceptible	71	16 (22.5)	1 (1.4)	24 (33.8)	3 (4.2)	10 (14.1)	7 (9.9)	9 (12.6)	1 (1.4)	0
Resistant	74	7 (9.5)	10 (13.5)	35 (47.3)	0	13 (17.6)	1 (1.3)	4 (5.4)	3 (4.1)	1 (1.3)
Blood samples	99	3 (3)b	11 (11.1)	65 (65.7)c	0	9 (9.1)	7 (7.1)	4 (4)	0	0
Susceptible	9	0	1 (11.1)	6 (66.7)	0	2 (22.2)	0	0	0	0
Resistant	90	3 (3.4)	10 (11.1)	59 (65.5)d	0	7 (7.8)	7 (7.8)	4 (4.4)	0	0

^aSusceptible indicates susceptibility to all antimicrobials tested. Resistant indicates resistance to at least one of the antimicrobials tested.

^bStatistical differences (P = 0.002) between A phylogroup strains from fecal and blood samples.

^cStatistical differences (P < 0.001) between B2 phylogroup strains from fecal and blood samples.

^dStatistical differences (P < 0.001) between resistant B2 phylogroup strains from fecal and blood samples.

FIG 3.

Frequency of virulence factors (VFs) analyzed in E. coli from fecal and blood sample strains and percentages of E. coli strains according to the number of VFs they carry. Adhesins include fimH (mannose-specific adhesin of type 1 fimbriae) and hra (heat-resistant agglutinin); siderophores include fyuA (yersiniabactin), iutA (aerobactin), and iroN (salmochelin receptor); toxins include hlyD (α-hemolysin), hlyF (hemolysin F), cnf1 (cytotoxic necrotizing factor 1), and clbB and clbN (colibactin); and the miscellaneous VF genes include iss (surface exclusion serum survival protein), traT (serum resistance), ompT (outer membrane protease), ibeA (invasion of brain endothelium), and usp (uropathogenic-specific protein). *, Statistical differences (P < 0.05) between each population, fecal and blood samples.

All VFs studied were detected in both populations. As expected, the clinical strains had a higher diversity (9 to 15 VFs) and frequency of VFs compared to those of the fecal samples (1 to 8 VFs) (Fig. 3). Finally, as supported by other authors (48), an association between some VFs (fyuA, iutaA, hlyF, iss, and traT) and strains carrying F plasmids was determined (P ≤ 0.003, Bonferroni’s correction was applied).

In conclusion, new information is provided about the plasmid background in strains isolated in a nonhospital setting. Although a similar trend was observed in the Inc groups from both populations, IncL/M, IncA/C, and IncN plasmids were only detected in clinical strains, whereas HI1 was only present in fecal strains. Also, two different evolutionary pathways followed by plasmids were observed as follows: specific IncI1 plasmids, such as IncI1 ST12, seem to have evolved by acquiring persistence, adaptive, and antibiotic resistance features relevant in clinical settings, whereas the more widespread multireplicon IncF plasmids have randomly acquired resistance genes. Additionally, the findings from this study confirm that strains from healthy individuals have less antimicrobial resistance and fewer VFs and display a higher diversity of phylogenetic lineages (in E. coli) than strains causing infection.